Agent Inbox

The inbox is organised as conversation threads, one per exchange between an agent and your team. A thread carries:

• The agent it belongs to and its type — SEO, content, paid social, email marketing, social studio, audience intelligence, or Ennie AI.
• The channel it arrived on: WhatsApp, email, SMS, or web chat.
• Whether the agent started it or a person did.
• A status of active, resolved, or archived, plus an unread count.
• The campaign it relates to, when there is one, so campaign work stays grouped.
• Any assets the agent produced — images, video, documents, ad creative, or campaign reports.

The inbox opens on three tabs — Conversations, My Agents, and Analytics — with summary counts above them. On Conversations you can narrow the list by:

• A free-text search across threads.
• Agent type, thread status, and source.
• An Unread Only toggle, for triaging quickly.
• One or more campaigns, using the campaign selector, when you want everything relating to a single push.

Selecting a thread opens it beside the list, so you keep your place.

A thread reads like a conversation, with messages from the agent, from you, and from teammates. From inside it you can:

1. Reply. Send a message back to the agent. Mentioning a teammate pulls them in.
2. Invite people. Add participants as owner, collaborator, or viewer, each with their own notification setting and their own read state.
3. Act on assets. Anything the agent produced is previewed in the thread with an approval status of pending, approved, or rejected. Approving marks it approved and records who approved it; rejecting asks for feedback, which is returned to the agent rather than discarded.
4. Close it out. Mark the thread read, or archive it when the work is done.

Some actions are too consequential to sit in a message thread, so they become approval requests instead. These cover publishing or creating a campaign, increasing a campaign budget, allocating budget, authorising a payment, publishing or modifying content, connecting or disconnecting an integration, deploying, changing, or deleting an agent, and exporting or deleting data. Each request carries:

• A summary of what is being asked, and a preview where one applies.
• A risk level of low, medium, high, or critical, with estimated cost and estimated reach where relevant.
• Who requested it, who it is assigned to, and an SLA deadline — a request past its deadline is flagged as breached.
• A status: pending, approved, rejected, expired, cancelled, or escalated.

Approving takes optional comments. Rejecting requires a reason, which goes back to the requesting agent. Pending approvals surface on the Dashboard and on the Campaigns screen as well as in the queue, and agency accounts get one combined queue across every client.

The inbox has an Update Mode control and a manual Refresh button, so you can choose how aggressively it updates:

• Auto — live updates, falling back to polling if the live connection is unavailable. This is the recommended setting.
• Polling only — refresh every thirty seconds.
• Manual only — nothing updates until you press Refresh.

Notifications outside the app are configured separately: email with a delivery frequency, WhatsApp after verifying your number with a code, and quiet hours to suppress messages between a start and end time.

1. Filter to unread, and work down.
2. Deal with high and critical risk approvals first — those are the ones with an SLA attached.
3. Approve or reject assets in the thread rather than replying in prose, so the agent gets a structured answer it can act on.
4. Reject with a real reason. The feedback is what improves the next attempt.
5. Archive finished threads so the active list stays a to-do list rather than a history.